Go on, spoil the mystique :)

Even fixing up the out-by-two error s/1+pos($$s)/pos($$s)-1/, I can't see how to correct the pos memory factor without explicitly resetting it? Which makes taking the reference and localising it redundant?

#! perl -slw
use strict;

sub tye {
    my $s= \shift @_;
    my( $p, $n )= @_;
    local( pos($$s) );
    0 while $$s =~ /\Q$p/g  &&  0 < --$n;
    return if  0 < $n;
    return 1+pos($$s);
}

my $data = 'a' x 100;

print "want: got";
printf "%3d : %3d\n", $_-1, tye( $data, 'a', $_ ) for 1 .. 30;

__END__
C:\test>junk
want: got
  0 :   2
  1 :   4
  2 :   7
  3 :  11
  4 :  16
  5 :  22
  6 :  29
  7 :  37
  8 :  46
  9 :  56
 10 :  67
 11 :  79
 12 :  92
Use of uninitialized value in printf at C:\test\junk.pl line 16.
 13 :   0
 14 :  16
 15 :  32
 16 :  49
 17 :  67
 18 :  86
Use of uninitialized value in printf at C:\test\junk.pl line 16.
 19 :   0
 20 :  22
 21 :  44
 22 :  67
 23 :  91
Use of uninitialized value in printf at C:\test\junk.pl line 16.
 24 :   0
 25 :  27
 26 :  54
 27 :  82
Use of uninitialized value in printf at C:\test\junk.pl line 16.
 28 :   0
 29 :  31
[download]

About the best performing modification of it I came up with is:

sub tyem {
    my( $s, $c, $n ) = @_;
    0 while $s =~ /\Q$c/g  &&  0 < --$n;
    my $rv = pos( $s )-1;
    pos($s) = -1;
    return if 0 < $n;
    return $rv;
}
[download]

which doesn't fair so well, though it is still quicker than any of the other regex-based solutions.

If you copy the string (and then destroy the copy), then it is a waste of time to set pos($s)= -1.

It was not an off-by-two error as I consider returning -1 (a 'true' value) for 'not found' to rather suck as an interface, and prefer to avoid returning a false value (0) for 'found at 1st character' so I meant to return the position of the character as asked, not the offset to it.

- tye